`
`Short Sellers and Financial Misconduct
`
`JONATHAN M. KARPOFF and XIAOXIA LOU∗
`
`ABSTRACT
`We examine whether short sellers detect firms that misrepresent their financial state-
`ments, and whether their trading conveys external costs or benefits to other investors.
`Abnormal short interest increases steadily in the 19 months before the misrepresen-
`tation is publicly revealed, particularly when the misconduct is severe. Short selling
`is associated with a faster time-to-discovery, and it dampens the share price inflation
`that occurs when firms misstate their earnings. These results indicate that short sell-
`ers anticipate the eventual discovery and severity of financial misconduct. They also
`convey external benefits, helping to uncover misconduct and keeping prices closer to
`fundamental values.
`
`SHORT SELLING IS A CONTROVERSIAL ACTIVITY. Detractors claim that short sell-
`ers undermine investors’ confidence in financial markets and decrease market
`liquidity. For example, a short seller can spread false rumors about a firm
`in which he has a short position and profit from the resulting decline in the
`stock price.1 Advocates, in contrast, argue that short selling facilitates market
`efficiency and the price discovery process. Investors who identify overpriced
`firms can sell short, thereby incorporating their unfavorable information into
`market prices. In his account of short selling in Allied Capital, Inc., hedge fund
`
`∗Foster School of Business, University of Washington and Lerner School of Business and Eco-
`nomics, University of Delaware, respectively. We thank especially Jerry Martin, who maintains
`the Karpoff-Lee-Martin database used in this study, and also Anup Agrawal, Uptal Bhattacharya,
`Hemang Desai, Karl Diether, Avi Kamara, Adam Kolansinski, Jennifer Koski, Srinivasan Krish-
`namurthy, Paul Laux, Paul Malatesta, Charu Raheja, Ed Rice, Ronnie Sadka, Katsiaryna Salavei,
`Mark Soliman, Ingrid Werner, two Journal of Finance referees, the Associate Editor, Campbell Har-
`vey, and seminar participants at the 2008 CRSP Forum, Concordia University, Yale Law School,
`Binghamton University–SUNY, Rutgers University, Syracuse University, Temple University, Uni-
`versity of Indiana, University of Washington, Vanderbilt Law School, and the California Corporate
`Finance Conference for helpful comments. We also thank the Q Group, The CFO Forum, and the
`Foster School of Business for financial support.
`1There are many anecdotes about such strategies, which former SEC Chairman Christopher
`Cox called “distort and short” (see “What the SEC really did on short selling,” The Wall Street
`Journal, July 24, 2008, A15). In 2000, for example, investor Mark Jakob turned a $241,000 profit
`by shorting Emulex stock and spreading an Internet rumor that Emulex’s CEO was stepping
`down amid an SEC investigation (see http://www.sec.gov/litigation/litreleases/lr16747.htm and
`http://www.sec.gov/litigation/litreleases/lr16857.htm). Leinweber and Madhavan (2001) report a
`case in which investors shorted Sea World stock and spread false rumors that Shamu, Sea World’s
`main attraction, was ill. For other examples, see Alistair Barr, “Short sellers: The good, the bad
`and the ugly,” MarketWatch, June 13, 2006.
`
`1879
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`1880
`
`The Journal of Finance R(cid:3)
`
`manager David Einhorn argues that short sellers even help uncover financial
`reporting violations (Einhorn (2008)).2
`In this paper, we investigate whether short sellers do in fact identify over-
`priced firms, and whether in the process they convey external benefits or harm
`upon other investors. To do so, we measure short selling in a set of firms that,
`ex post, clearly were overpriced: those that are disciplined by the SEC for fi-
`nancial misrepresentation. In our sample of 454 firms from 1988 through 2005,
`96% have negative abnormal returns on the days their misconduct was pub-
`licly revealed, with an average 1-day stock price decline of 18.2%. These firms
`therefore provide a natural test of the view that short sellers can anticipate
`bad news.
`The results of three tests indicate that short sellers are proficient at iden-
`tifying financial misrepresentation before it becomes public. First, abnormal
`short interest rises significantly in the 19-month period before the misrepre-
`sentation is publicly revealed. Second, the amount of short selling is positively
`related to measures of misconduct severity, indicating that short sellers take
`larger positions when the misrepresentation is particularly egregious. And
`third, short interest-based indicators of financial misrepresentation in any
`given firm-month are significantly related to the actual presence of misrep-
`resentation, as revealed in subsequent SEC documents.
`We also investigate whether short selling has external effects on other in-
`vestors. We do not find evidence that short selling imposes external harm by
`triggering a cascade of selling when the misconduct is publicly revealed. To
`the contrary, short selling conveys positive externalities to other investors, in
`two ways. First, the amount of prior short selling is positively related to how
`quickly the misconduct is publicly revealed. Our point estimates indicate that,
`among firms that are 12 months into their misrepresentation, those with ab-
`normal short interest at the 75th percentile will be publicly revealed 8 months
`before firms at the 25th percentile.
`Second, short selling dampens the amount by which prices are inflated while
`firms report incorrect financial statements. This improves price efficiency and
`decreases the transfer from uninformed investors who buy shares from insiders
`or the firm before the misconduct is revealed to the public. We estimate that
`this price impact translates into savings for uninformed investors of around
`1.67% of the firm’s market capitalization on average. Some of these savings
`are captured by short sellers, who earn profits that average 0.58% of equity
`value. Even net of such profits, the average net external benefit to uninformed
`investors equals 1.09% of the firm’s equity value.3
`These findings do not address whether short selling in general is informed
`and beneficial for other investors. For example, we cannot rule out the possibil-
`ity that some short sellers are noise traders, or that some seek to manipulate
`prices through false rumors. But in our events—in which company managers
`
`2Lamont (2004) and Jones and Lamont (2002) summarize the debate over whether short selling
`fosters market efficiency or facilitates harmful manipulation. See also Wilchins (2008).
`3These point estimates correspond to our first measure of abnormal short interest, ABSI(1).
`Depending on the specific measure, our point estimates of the net external benefit range from
`0.19% to 1.53% of equity value. See Section V.C and Table IX below.
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`Short Sellers and Financial Misconduct
`
`1881
`
`produce falsified financial statements—short sellers play a significant role in
`identifying, uncovering, and mitigating the effects of financial misconduct.
`This paper is organized as follows. In Section I, we review related research
`and argue that our sample and test design are uniquely well suited to exam-
`ine whether short sellers anticipate and help uncover financial misconduct.
`Section II describes our data and measures of abnormal short interest.
`Section III reports on tests of short sellers’ ability to anticipate financial miscon-
`duct, and Section IV examines short sellers’ external effects on other investors.
`Section V concludes.
`
`I. Related Research
`Our investigation is related to a large body of research that examines whether
`short sellers target overvalued stocks.4 The results are somewhat mixed.
`Asquith and Meulbroek (1996) and Desai et al. (2002) find that stocks that
`are highly shorted in one month tend to underperform in the next month, and
`Diether, Lee, and Werner (2009) find that short sellers appear to take advan-
`tage of short-term overreaction in stock prices. Christophe, Ferri, and Angel
`(2004), Christophe, Ferri, and Hsieh (2009), and Liu, Ma, and Zhang (2008)
`find that short selling increases before negative earnings announcements, an-
`alyst downgrades, and mortgage loss-related write-downs. In contrast, Daske,
`Richardson, and Tuna (2005) do not find any predictive ability of short selling,
`and Henry and Koski (2010) find no evidence of informed short selling around
`SEO announcements.
`Our empirical tests employ measures of abnormal short interest that condi-
`tion on firm characteristics, and thus are related to inquiries into whether short
`sellers use information about firm fundamentals. Dechow et al. (2001), Asquith
`et al. (2005), and Duarte, Lou, and Sadka (2006) find that short interest is re-
`lated to market capitalization, book-to-market, and momentum. Richardson
`(2003) fails to find evidence that short sellers target firms with high accruals.
`But Cao, Dhaliwal, and Kolasinski (2006) find that short sellers do target firms
`with high accruals after controlling for surprises in earnings announcements.
`We find that short interest is related to accruals, as well as market capitaliza-
`tion, book-to-market, momentum, insider selling, institutional ownership, and
`share turnover.
`Three prior studies are most closely related to ours. Dechow, Sloan, and
`Sweeney (1996) report an increase in short interest in the 2 months before
`an SEC release in a sample of 27 Accounting and Auditing Enforcement Re-
`leases. Desai, Krishnamurthy, and Venkataraman (2006) and Efendi, Kinney,
`and Swanson (2006) examine short selling before the accounting restatements
`in a database compiled by the Government Accountability Office (GAO).5 Our
`investigation differs from these papers in several ways. First, we introduce
`
`4See Figlewski (1981), Asquith and Meulbroek (1996), Desai et al. (2002), and Asquith, Pathak,
`and Ritter (2005).
`5For a description of the GAO restatement data, see http://www.gao.gov/special.pubs/gao-06-
`1079sp/.
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`1882
`
`The Journal of Finance R(cid:3)
`
`several controls for the severity of the misconduct, allowing us to infer whether
`short selling affects stock prices directly, or whether it merely serves as a proxy
`for misconduct severity. Second, we examine whether short selling tends to con-
`centrate in the misconduct firms. And third, we estimate the external effects on
`uninformed investors—including whether short selling helps expose financial
`misconduct and whether it dampens price inflation during the violation period.
`The Internet Appendix contains a tabular summary of the results that are new
`to this paper.6
`The data we use also provide for more powerful tests than the GAO restate-
`ment data. Hennes, Leone, and Miller (2008) report that 76% of the restate-
`ments in the GAO database are simple errors rather than misrepresentation
`or fraud, a concern also expressed by Files, Swanson, and Tse (2009). This sug-
`gests that the GAO database contains a large number of misclassified events.
`Even when restatements do reflect financial misconduct, they can occur many
`months after the misconduct is public knowledge. In our sample, SEC inquiries
`into financial misconduct are resolved 41 months after the initial public rev-
`elation, on average. Using a restatement that is made during or after that
`41-month period would misclassify when the misrepresentation was or was
`not public knowledge.
`
`II. Data and Short Interest Measures
`A. Financial Misrepresentation Data
`
`To avoid the data problems discussed above, we use the Karpoff, Lee, and
`Martin (2008a, 2008b) (hereafter KLM) database to identify all 632 SEC en-
`forcement actions for financial misrepresentation initiated from 1988 through
`2005.7 These data identify the period during which the misrepresentation oc-
`curred and also the trigger event, which is the initial public revelation of the
`misconduct. This allows us to focus on short selling around the initial public
`revelation. Short interest data are available for 474 of the 632 firms, and 454
`firms have sufficient data on CRSP to calculate returns on their revelation
`dates.
`To illustrate the nature of our data and tests, it is useful to review the se-
`quence of events that constitute an SEC enforcement action.8 These events are
`summarized in Figure 1. Most enforcement actions follow a conspicuous trigger
`
`6The Internet Appendix is available on The Journal of Finance website at http://www.afajof.
`org/supplements.asp.
`7Karpoff et al. (2008a, p. 10) report that the database is collected from “ . . . Lexis-Nexis’ FED-
`SEC:SECREL library for information on SEC securities enforcement actions, the FEDSEC:CASES
`library for information on litigated enforcement actions, and the Academic Business News, General
`News, and Legal Cases libraries for news releases (frequently issued by defendant firms) about
`each enforcement action . . . the SEC’s website at http://www.sec.gov, which contains all SEC public
`releases relating to enforcement actions since September 19, 1995 . . . the Department of Justice
`itself, which provided . . . further data on enforcement outcomes [, and] the Department of Justice’s
`Corporate Fraud Task Force website at http://www.usdoj.gov.”
`8The following two paragraphs follow Section III in Karpoff et al. (2008b).
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`Short Sellers and Financial Misconduct
`
`1883
`
`Violation Period
`
`*
`
`Enforcement Period
`
`Regulatory Period
`
`Violation
`Begins
`
`Violation
`Ends
`
`Trigger
`Event
`
`Inquiry
`Event
`
`Investigation
`Event
`
`Wells
`Notice
`
`Proceedings Events
`Concluding
`Regulatory
`Proceeding
`
`Initial
`Regulatory
`Proceeding
`
`Violation to Revelation
`Enforcement Events
`* The initial filing of a private lawsuit usually occurs soon after the trigger event.
`
`Figure 1. Timeline of a typical enforcement action.
`
`event that publicizes the potential misconduct and attracts the SEC’s scrutiny.
`Common trigger events include self-disclosures of malfeasance, restatements,
`auditor departures, and unusual trading. Here are two examples of trigger
`events from our sample:
`
`1. On November 21, 2000, Lucent Technologies Inc. announced that it had
`identified a revenue recognition issue in its already-reported fourth quar-
`ter report as the company was completing its financial statements for fiscal
`year 2000. The company also told investors not to rely on its first-quarter
`forecast of 2001. Share prices fell 16% on the announcement day.
`2. On November 13, 2003, Virbac Corporation announced that it was delay-
`ing the filing of its third-quarter 2003 Form 10-Q. Share prices fell 22%
`on the announcement day.
`
`Following a trigger event, the SEC gathers information through an informal
`inquiry that may develop into a formal investigation of financial misconduct.
`At this point the SEC may drop the case, in which case it does not appear in our
`sample. If the SEC proceeds, it typically sends a Wells Notice to prospective
`defendants, notifying them that it intends to begin enforcement proceedings.
`It then imposes administrative sanctions and/or seeks redress through civil
`actions. Some cases are referred to the Department of Justice and lead to
`criminal charges as well. The SEC releases its findings and penalties in its
`Administrative Proceedings and Litigation Releases, and every enforcement
`action in our sample has at least one such release. These releases provide
`detailed information on the period over which the misrepresentation occurred—
`which we label the violation period—as well as other information that we use
`in our empirical tests.
`As reported in Table I, the events illustrated in Figure 1 typically take several
`years to play out. In our sample of enforcement events, the median length of
`the violation period is 24 months, and the median length from the beginning of
`the violation until its initial public revelation is 26 months. The period from the
`initial public revelation until the end of the enforcement action is an additional
`41 months. Table I shows that the number of enforcement actions, the median
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`1884
`
`The Journal of Finance R(cid:3)
`
`Table I
`Description of the Financial Misrepresentation Sample
`This table describes the yearly distribution of the 632 SEC enforcement actions for financial
`misrepresentation from 1988 through 2005. The violation period is the date the financial misrep-
`resentation began until it ended, as identified in SEC litigation or administrative releases. The
`revelation date is the earliest date that information about the misrepresentation was made pub-
`lic. Most revelation dates are identified in SEC releases, and the remaining are identified in the
`Karpoff et al. (2008a, 2008b) database. Revelation events include firm disclosures, restatements,
`auditor changes, SEC filing delays, whistle-blower charges, class action lawsuit filings, bankruptcy
`filings, and SEC actions (informal inquiries, formal investigations, Wells Notices, or first regulatory
`proceeding).
`
`Violation Period
`(Months)
`
`Violation Beginning
`to Public Revelation
`(Months)
`
`Number of Cases
`
`Mean
`
`Median
`
`Mean
`
`Median
`
`25
`13
`28
`34
`35
`32
`46
`29
`37
`34
`36
`36
`67
`49
`70
`31
`17
`13
`632
`
`20
`25
`20
`34
`26
`24
`22
`26
`29
`26
`30
`33
`27
`26
`32
`32
`32
`42
`28
`
`24
`23
`15
`24
`24
`20
`17
`24
`27
`24
`21
`30
`24
`21
`24
`33
`32
`36
`24
`
`31
`30
`29
`35
`32
`31
`29
`28
`33
`32
`33
`33
`28
`25
`33
`35
`41
`46
`31
`
`28
`23
`22
`30
`28
`24
`22
`25
`27
`24
`25
`24
`23
`20
`27
`29
`36
`36
`26
`
`Year
`
`1988
`1989
`1990
`1991
`1992
`1993
`1994
`1995
`1996
`1997
`1998
`1999
`2000
`2001
`2002
`2003
`2004
`2005
`Total
`
`violation period, and median period from the beginning of the violation to its
`public revelation generally increased from 1988 to 2005.
`Panel A of Table II reports that news about financial misrepresentation is
`associated with large declines in stock price. Return data are available for 454
`of our sample firms. For 359 of these firms, the trigger event reported in the
`KLM database is identified in the SEC’s administrative and litigation releases.
`The mean 1-day market-adjusted return on the SEC-identified trigger date is
`−20.7%, and the median is −15.0%.
`For 95 of our events the SEC identified no trigger date, or the KLM database
`indicates that there was an earlier public revelation of the misconduct. In 37
`of our events, for example, the start of a class action lawsuit is the earliest
`public revelation of the misconduct. The mean 1-day market-adjusted return
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`Short Sellers and Financial Misconduct
`
`1885
`
`Table II
`Share Price Reactions to Announcements of Financial
`Misrepresentation
`This table presents summary statistics on the 1-day market-adjusted returns for key dates in the
`sample of 454 SEC enforcement actions for financial misrepresentation from 1988 to 2005, for which
`sufficient returns data are available on CRSP. The market-adjusted return is the firm’s return
`minus the CRSP value-weighted return on the same day. Panel A reports market-adjusted returns
`for the revelation date, which is the earliest date that information about the misrepresentation
`was revealed to the public. Most (359) revelation dates are identified by the SEC, and include firm
`disclosures, restatements, auditor changes, SEC filing delays, and whistle-blower charges. In 95
`cases, the initial revelation date is identified in the Karpoff et al. (2008a, 2008b) database. These
`events include announcements of an SEC informal inquiry or formal investigation, announcements
`of a Wells Notice, the initiation of regulatory proceedings, class action lawsuits, and bankruptcy
`announcements. Panel B reports on important announcements about the misrepresentation that
`were made after the public revelation date. These subsequent events include announcements of an
`informal SEC inquiry, formal SEC investigation, Wells Notice, initiation of regulatory proceedings,
`initiation of class action lawsuits, and bankruptcy. There are a total of 844 such subsequent
`announcements. Of these, 371 are follow-ups to the initial revelation date. Of these 371 cases, 274
`have a third announcement, 147 have a fourth announcement, and 46 have a fifth announcement.
`
`N
`
`Mean (%)
`
`Median (%)
`
`t-Stat
`
`All initial revelation dates
`SEC-identified trigger event
`Other initial revelation events
`–SEC informal inquiry
`–SEC formal investigation
`–SEC Wells Notice
`–Regulatory proceedings begin
`–Class action lawsuits begin
`–Bankruptcy
`
`Panel A: Initial Public Revelation Date
`−18.20
`−20.70
`−8.90
`−12.10
`−9.32
`−1.03
`−6.29
`−5.93
`−20.40
`
`454
`359
`95
`15
`22
`1
`12
`37
`8
`
`2nd announcement
`3rd announcement
`4th announcement
`5th announcement
`6th or higher announcement
`All subsequent announcements combined
`
`Panel B: Important Subsequent Announcements
`−9.61
`−7.22
`−3.52
`−0.00
`−13.76
`−7.28
`
`371
`274
`147
`46
`6
`844
`
`−11.10
`−15.00
`−5.77
`−11.70
`−6.09
`−1.03
`−1.97
`−3.73
`−14.40
`
`−4.96
`−3.97
`−1.95
`−0.90
`−6.09
`−3.69
`
`−19.90
`−19.00
`−8.55
`−5.17
`−4.62
`N/A
`−2.98
`−5.12
`−3.00
`
`−12.41
`−8.85
`−4.88
`0
`−1.53
`−15.30
`
`for these 37 cases is −5.93%. Other less common revelation dates include the
`announcement of a formal SEC investigation (22 events), an informal SEC
`inquiry (15), the initial regulatory action and SEC release (12), and bankruptcy
`filing (8). For all 95 of the revelation dates that are not identified by the SEC,
`the mean 1-day market-adjusted return is −8.9%.
`Averaging over all 454 initial revelation dates, the mean abnormal return
`is −18.2% and the median is −11.1%. In the tests that follow we use data
`from all 454 events. The results are qualitatively identical, however, if we
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`1886
`
`The Journal of Finance R(cid:3)
`
`limit the sample to the 359 events for which the SEC identified the trigger
`date. Either way, these results indicate that public announcements that firms
`violated financial reporting rules are associated with large declines in share
`values. These are exactly the type of events that benefit short sellers.
`Panel B of Table II reveals that share prices tend to decrease further when
`additional news about the misrepresentation is revealed to the public. The an-
`nouncements in this panel include SEC informal inquiries, SEC formal investi-
`gations, Wells Notices, the initiation of regulatory proceedings, the initiation of
`class action lawsuits, and bankruptcies. A total of 371 of the 454 events have a
`second announcement. The mean 1-day return for these 371 second announce-
`ments is −9.6%. A total of 274 events have a third announcement, with a mean
`1-day return of −7.2%. Combining all 844 subsequent announcements in Panel
`B, the mean 1-day return is −7.3% with a t-statistic of −15.3. These numbers
`indicate that subsequent information about these firms’ financial misconduct—
`even after the initial public revelation—also tends to be unfavorable.
`
`B. Short Interest Data and Related Data
`
`Our tests examine the ability of short sellers to depict misrepresentation
`before it is publicly revealed, and thus we focus on short interest during the
`violation period immediately before the initial public revelation dates that are
`summarized in Panel A of Table II. Monthly short-interest data are obtained
`from the New York Stock Exchange (NYSE), the American Stock Exchange
`(Amex), and NASDAQ for the period January 1988 to December 2005.9 Short
`interest reflects the open short positions of stocks with settlements on the last
`business day on or before the 15th of each calendar month. Settlement, however,
`takes a few days, and for a short sale transaction to be recorded in month t, it
`must occur before or on the trade date. Before June 1995, the trade date was 5
`days before the settlement date; currently, it is 3 days before. We define month
`t as starting the day after the trade date of calendar month t−1 and ending
`on the trade date of calendar month t. Raw short interest for firm i in month
`t, SIit, is the percent of total shares outstanding in month t. The pooled mean
`level of SIit over all months for all firms covered by the short interest data is
`1.65%.
`Monthly stock returns and market capitalization are constructed from daily
`data obtained from CRSP using the month definition explained above. Some of
`the analysis requires data on past returns and institutional ownership. Con-
`sequently, we use CRSP data from January 1987 through December 2005. We
`obtain data on institutional ownership from the CDA/Spectrum database pro-
`vided by Thomson Financial. The data, derived from institutional investors’
`
`9Daily data from January 1, 2005 through August 6, 2007 recently have become available to
`researchers. These data, however, cover only a small number of the enforcement actions in our
`sample. The daily data also do not contain information about short positions that are covered,
`making it impossible to compute net changes in short interest. The monthly data therefore are
`well suited to our tests.
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`Short Sellers and Financial Misconduct
`
`1887
`
`quarterly filings of SEC Form 13F, include quarterly holdings for each stock
`for each quarter between December 1987 and December 2005.
`
`C. Abnormal Short Interest
`
`(1)
`
`In addition to raw short interest, we examine three measures of abnormal
`short interest. For firm i in month t, abnormal short interest equals
`ABSI( j)it = SIit − E(SI( j)it), j = 1, 2, 3,
`where SIit is raw short interest and E(SI(j)it) is the expected short interest
`based upon one of three benchmarks j that reflect the firm’s characteristics.
`The first benchmark, E(SI(1)it), controls for the firm’s market capitalization,
`book-to-market ratio, past stock performance, and industry. These controls
`reflect findings by Dechow et al. (2001), Asquith et al. (2005), and Duarte et
`al. (2006) that short interest is related to market capitalization, the book-to-
`market ratio, and momentum. At the beginning of each month, each stock is
`assigned to one of 27 portfolios constructed by independently sorting stocks
`by size, book-to-market, and momentum, all measured at the end of the prior
`month. Each of the 27 portfolios is further partitioned into industry groups
`using two-digit SIC codes. We exclude the sample firms in constructing the
`matching portfolios.
`In particular, E(SI(1)it) is the fitted value from a cross-sectional regression
`that is estimated for each month t:
`
`SIit = medium(cid:2)
`g=low
`
`sgtSizeigt + medium(cid:2)
`g=low
`
`bgt BMigt + medium(cid:2)
`g=low
`
`mgt Momigt + K(cid:2)
`k=1
`
`φkt Indikt + uit.
`(2)
`
`The first three sets of explanatory variables are dummy variables that jointly
`define the 27 size-, book-to-market-, and momentum- based portfolios. For ex-
`ample, if firm i is assigned to the portfolio with the lowest market capitalization
`in month t, thenSize i,low,t, = 1, Sizei,medium,t, = 0, and Sizei,high,t = 0. Industry
`dummy Indikt = 1 if firm i belongs to industry k in month t, and K is the total
`(cid:3)K
`k=1 Indikt= 1 (so the intercept term
`number of industries. By construction,
`is omitted). Each monthly regression uses all firms listed on NYSE, Amex, or
`NASDAQ that are not in our SEC enforcement action sample and for which
`data on short interest, market capitalization, book-to-market, and momentum
`are available over the period 1988 through 2005.
`Table III reports the time-series averages of the coefficient estimates (ex-
`cluding industry dummies) of the monthly cross-sectional regressions. The
`associated t-statistics are computed with Newey–West (1987) corrections for
`serial correlation using three lags. The base portfolio in this regression is the
`portfolio with the highest market capitalization, book-to-market ratio, and mo-
`mentum for each industry. This means that the coefficients are interpreted as
`the difference between the short interest of the given portfolio and that of the
`base portfolio. The results show that the largest firms have the highest short
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`1888
`
`The Journal of Finance R(cid:3)
`
`Table III
`Models Used to Calculate Abnormal Short Interest
`For each month t, short interest (SI) is regressed on variables that are likely to explain the level
`of short interest that is unrelated to short sellers’ information about financial misconduct. Short
`interest (SI) is the number of shares shorted as a percentage of the number of shares outstanding.
`The table reports the time-series means and t-statistics of the monthly coefficient estimates. For
`Model 1:
`
`bgt BMigt + medium(cid:2)
`g=low
`
`φkt Indikt + uit.
`
`SIit = medium(cid:2)
`g=low
`
`sgt Sizeigt + medium(cid:2)
`g=low
`
`mgt Momigt + K(cid:2)
`k=1
`Explanatory variables include size, the book-to-market ratio, and momentum, all measured at the
`beginning of month t. The explanatory variables are dummy variables. For example, if firm i is
`assigned to the portfolio with the lowest market capitalization in month t, then Sizei,low,t = 1,
`Sizei,medium,t = 0, and Sizei,high,t = 0. Model 2 includes dummy variables for share turnover and
`institutional ownership, and Model 3 includes continuous variables for total accruals and insider
`selling. All three regressions include industry dummies with Indikt = 1 if firmi belongs to industry
`k in month t. Variable K is the total number of industries, and industry is defined using two-digit
`SIC codes from CRSP. The sample includes all firms listed on NYSE, Amex, or NASDAQ that are
`not in the SEC enforcement action sample and for which data are available during the 1988 to
`2005 period. t-statistics are computed with Newey–West (1987) corrections for serial correlation
`using three lags.
`
`Sizelow
`
`Sizemedium
`
`BMlow
`
`BMmedium
`
`Momentumlow
`
`Momentummedium
`
`Turnoverlow
`
`Turnovermedium
`
`Institutional ownershiplow
`
`Institutional ownershipmedium
`
`Total accruals
`
`Insider selling
`
`Industry controls
`Adj-R2
`
`Model 1
`(Used to
`Calculate
`ABSI(1))
`−1.952
`[−13.09]
`−0.922
`[−9.92]
`0.345
`[7.49]
`−0.353
`[−14.12]
`0.402
`[8.16]
`−0.147
`[−5.48]
`
`Model 2
`(Used to
`Calculate
`ABSI(2))
`−0.709
`[−8.22]
`−0.322
`[−4.76]
`0.270
`[6.51]
`−0.266
`[−11.92]
`0.454
`[11.07]
`0.093
`[3.64]
`−2.261
`[−16.10]
`−1.899
`[−16.14]
`−0.949
`[−10.46]
`−0.588
`[−8.38]
`
`Yes
`0.21
`
`Yes
`0.27
`
`Model 3
`(Used to
`Calculate
`ABSI(3))
`−0.813
`[−8.32]
`−0.395
`[−5.02]
`0.264
`[6.28]
`−0.286
`[−12.05]
`0.466
`[10.30]
`0.093
`[3.59]
`−2.248
`[15.73]
`[−1.88]
`[−15.72]
`−0.931
`[−8.94]
`−0.531
`[−6.84]
`0.419
`[7.38]
`3.823
`[10.28]
`Yes
`0.28
`
`Coalition for Affordable Drugs IV LLC - Exhibit 1035
`
`
`
`Short Sellers and Financial Misconduct
`
`1889
`
`interest. Both the book-to-market ratio and momentum have U-shaped rela-
`tions with short interest, as indicated by the different signs of blow and bmedium,
`and mlow and mmedium. The relation between the book-to-market ratio and short
`interest is consistent with the finding in Dechow et al. (2001). The U-shaped
`relation between short interest and momentum also is documented by Duarte
`et al. (2006). Stocks with the lowest book-to-market ratios and lowest past
`performance are most highly shorted.
`Our second measure of abnormal short interest, ABSI(2)it, includes addi-
`tional controls for share turnover and institutional ownership, which D’Avolio
`(2002) shows are related to short sales constraints. The coefficients reported
`in the second column of Table III indicate that short interest increases with
`both share turnover and institutional ownership. The fitted values from each
`monthly cross-sectional regression are used to estimate E(SI(2)it), the expected
`amount of short interest for firm i in month t, which is in turn used to calculate
`ABSI(2)it = SIit − E(SI(2)it).
`Our third measure of abnormal short interest, ABSI(3)it, expands the number
`of control variables to include total firm accruals and insider selling. Healy
`(1985), Dechow et al. (2010), and others show that accruals can be used to
`manipulate earnings, and Agrawal and Cooper (2008) show that insider selling
`is correlated with financial misconduct at many firms. Einhorn (2008) reports
`that many short sellers base their positions on accruals and insider selling even
`in the absence of any specific knowledge about the firm. ABSI(3)it reflects short
`sellers’ information over and above their knowledge about accruals, insider
`selling, or the other control variables.
`Our measure of total accruals for firm i in month t is the same as that used
`by Richardson et al. (2005):
`
`Total accruals = WCit + NC Oit + F INit
`(Assetsit + Assetsi,t−12)/2
`
`.
`
`(3)
`
`Here, WCit is firm i’s change in noncash working capital. It is measured as
`the change in current operating assets net of cash and short-term investments,
`minus the change in current operating liabilities net of short-term debt. Non-
`current operating accruals, NCOit, is the change in noncurrent assets net of
`long-term nonequity investments and advances, less the change in noncurrent
`liabilities net of long-term debt. The change in net financial assets, FINit,
`is